KR20240079334A - Apparatus And Method For Removing Invalid Spectator Trajectory Fake Big Data - Google Patents

Abstract

The present invention relates to an apparatus and method for removing invalid visitor trajectory imitation big data, and more specifically, to obtain real trajectory data including content whose ID, time, and physical location in arbitrary space are determined. An acquisition unit, a fake data generator that generates fake data having the same column as the actual trajectory data using a deep learning model, and fake data that cannot be physically matched based on the actual trajectory data among the fake data. It relates to an apparatus and method for removing invalid visitor trajectory imitation big data including a filtering unit for filtering.

Description

Apparatus And Method For Removing Invalid Spectator Trajectory Fake Big Data}

The present invention relates to an apparatus and method for removing invalid visitor trace imitation big data.

Due to the increase in leisure time, interest in cultural life is increasing among people of all ages and genders, and as a result, various exhibitions are being held. Recently, exhibitions are providing a multifaceted viewing experience through the convergence of the Internet and smart technology in accordance with technological developments, and the quantitative and qualitative levels of exhibitions are improving in line with the increase in the cultural level of visitors, leading to gradual specialization and diversification of exhibitions. It is being done.

At this time, a lot of data is needed to improve user experience and operate exhibitions through integration with smart technology, and a lot of data becomes a dataset from which artificial intelligence models can be learned. At this time, the amount of data that can be acquired in reality is not enough for the artificial intelligence model to learn, so it is necessary to artificially increase the amount of data. In this process, inappropriate data may be included, and artificial intelligence models can be learned without any processing or correction. When an intelligence model is trained, incorrect results may be output repeatedly. Therefore, in this field of technology, there is an urgent need for technology that can generate, process, and correct the trajectory data of numerous visitors required for exhibition operation.

Republic of Korea Patent Registration No. 10-1902315 Republic of Korea Patent Publication No. 10-2022-0049165

The present invention is intended to solve the above problems, and has the same column as the actual trajectory data to create an artificial intelligence model that provides more accurate output values by processing and correcting big data that simulates the trajectories of numerous visitors required for exhibition operation. The purpose is to obtain an apparatus and method for removing invalid visitor trajectory fake big data that generates fake data and filters fake data containing at least one of time and content that cannot be physically matched to the fake data. Do this.

The technical problems to be achieved by the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned can be clearly understood by those skilled in the art from the description of the present invention.

In order to achieve the above object, the apparatus for removing invalid visitor trajectory imitation big data of the present invention is an actual trajectory data acquisition unit that acquires actual trajectory data including content whose ID, time, and physical location in an arbitrary space are determined. ; a fake data generator that generates fake data having the same columns as the actual trajectory data using a deep learning model; and a filtering unit that filters out fake data that cannot be physically matched with the actual trajectory data among the fake data.

In order to achieve the above object, the method for removing invalid visitor trajectory imitation big data of the present invention is to obtain actual trajectory data including content whose ID, time, and physical location in an arbitrary space are determined by the actual trajectory data acquisition unit. An actual trajectory data acquisition step in which is acquired; A fake data generation step in which fake data having the same column as the actual trajectory data is generated by a fake data generator using a deep learning model; and a filtering step in which, by a filtering unit, fake data that cannot be physically matched is filtered out of the fake data using the actual trajectory data as a standard.

As described above, according to the present invention, fake data having the same column as actual trajectory data is generated, and fake data containing at least one of time and content that cannot be physically matched to the fake data is filtered, thereby improving exhibition operation. There is a remarkable effect in processing and correcting the necessary big data to simulate the trajectories of numerous visitors.

In addition, the present invention has the effect of creating an artificial intelligence model that provides more accurate output values by learning an artificial intelligence model using processed and corrected trajectory imitation big data.

The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the detailed description and claims.

Figure 1 is a diagram illustrating the configuration of an apparatus for removing invalid visitor trajectory fake big data according to the present invention.
Figure 2 is a diagram showing valid and invalid fake data according to an embodiment of the present invention.
Figure 3 is a flowchart of a method for removing invalid visitor trajectory fake big data according to the present invention.

The terms used in this specification are general terms that are currently widely used as much as possible while considering the function in the present invention, but this may vary depending on the intention or precedent of a person working in the art, the emergence of new technology, etc. In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the description of the relevant invention. Therefore, the terms used in the present invention should be defined based on the meaning of the term and the overall content of the present invention, rather than simply the name of the term.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by a person of ordinary skill in the technical field to which the present invention pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and unless clearly defined in the present application, should not be interpreted in an ideal or excessively formal sense. No.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Figure 1 is a diagram illustrating the configuration of an apparatus for removing invalid visitor trajectory fake big data according to the present invention. Figure 2 is a diagram showing valid and invalid fake data according to an embodiment of the present invention. Figure 3 is a flowchart of a method for removing invalid visitor trajectory fake big data according to the present invention.

Device for removing invalid visitor trajectories and fake big data

First, looking at Figure 1, the apparatus for removing invalid fake visitor big data of the present invention includes an actual trajectory data acquisition unit 100, a fake data generation unit 200, and a filtering unit 300.

More specifically, the real trajectory data acquisition unit 100 acquires real trajectory data including content whose ID, time, and physical location in an arbitrary space are determined.

The ID referred to in the present invention may be at least one of unique numbers and letters given to recognize individual visitors. The time referred to in the present invention is the time when a visitor arrives at a physical location in a certain space. Depending on one embodiment, time may include units of years, months, days, hours, minutes, and seconds. The arbitrary space may be implemented as two-dimensional coordinates, and thus the physical location of content in the arbitrary space may be specified. In other words, the order of content can be confirmed according to a certain direction. And the real trajectory data (real data) is data actually obtained from a large number of visitors in any space where exhibitions that visitors can view, such as exhibitions or science museums, are held. An example of the real trajectory data (Real data) is shown in [Table 1] below.

ID hour contents One 220915:12:36 A One 220915:15:07 B 2 220915:11:10 C .
.
. .
.
. .
.
.

In addition, the actual trajectory data acquisition unit 100 is characterized by arranging contents of the actual trajectory data in chronological order by ID. This is because, when separately acquired data is listed in the order in which it was acquired, it is difficult to check the order and number of time-dependent content by ID, and it takes a considerable amount of time to check this, which reduces data processing speed and efficiency. It is to improve.

Next, the fake data generator 200 uses a deep learning model to generate fake data having the same columns as the actual trajectory data.

The synthetic data may include content with the same ID, time, and physical location in arbitrary space as the actual trajectory data. However, this is not data obtained from actual visitors in a random space, but data artificially generated using the deep learning model.

At this time, the fake data generator 200 is characterized in that the deep learning model is a CTGAN (Conditional Tabular GAN) model. In general, the GAN model is a network that artificially generates data similar to actual trajectory data including discrete data by utilizing mutually adversarial discriminators and generators, and is a model that belongs to unsupervised learning in machine learning. In other words, it is a model that complements the fact that the existing GAN model cannot be applied to discrete data.

Next, the filtering unit 300 filters fake data that cannot be physically matched with the actual trajectory data among the fake data.

The actual trajectory data referred to in the present invention has a time (t) value according to content (C), which is a specific location in arbitrary space. The fake data generated from CTGAN, a deep learning model in the fake data generator 200, may include invalid data that violates the characteristics of the trajectory data. If the fake time value corresponding to a specific position among the fake data properties is a value that cannot be generated when looking at the front-end correlation of the actual trajectory data, the filtering unit 300 determines that the fake data is invalid (Invalid). It can be removed as it is judged to be fake data or fake data that cannot be physically matched.

Specifically, the filtering unit 300 is valid if the time of any fake data is within the time range of any real trajectory data and the content of any fake data is within the content range of any real trajectory data. If the content of the time filter module 310 and the arbitrary fake data that are determined to be fake data exist within the content of the random real trajectory data and the time of the random fake data exists within the time range of the random real trajectory data, it is valid. ) It is characterized by including a content filter module 320 that determines it to be fake data.

Looking at the embodiment of FIG. 2, an arbitrary space may include all contents (C0-Cn) whose physical locations are determined. Additionally, the real trajectory data acquisition unit 100 may acquire real trajectory data including a plurality of contents (C0-Cn) dependent on time (t0-tn) targeting ID 1. That is, the actual trajectory data may include C0 content at time t0, C2 content at t1, C5 content at t2, Cn-1 content at time tn-1, and Cn content at tn. Additionally, the fake data generator 200 may generate fake data including a plurality of contents (Cy1-Cm) dependent on time (tx1-txm) targeting ID 1. Here, n and m are positive integers. That is, the fake data may include Cy1 content at time tx1 and Cy2 content at time tx2.

If the tx1 time is within the range from to to t1, Cy1 content can be C1 content, but cannot be one of C3 to Cn content. In other words, Cy1 content cannot be physically matched if it is one of C3 to Cn content. That is, according to one embodiment, the time filter module 310 determines that the fake data is invalid if the Cy1 content at time tx1 is C5 and is fake data that cannot be physically matched. The corresponding fake data can be removed. Therefore, the time filter module 310 can determine whether content is matched based on time.

Also, if Cy2 content is within the range of C2 content to C5 content, tx2 time must exist within the t1 to t2 time range. In other words, if the tx2 time is outside the t1 to t2 range, it cannot be physically matched. That is, according to one embodiment, the content filter module 320 determines that if the tx2 time of the Cy2 content does not exist in t1 to t2 but is t5, it is invalid and is fake data that cannot be physically matched. You can judge and remove the corresponding fake data. Therefore, the content filter module 320 can determine whether or not time matches based on the content.

Method for removing invalid visitor trajectories and fake big data

Referring to Figure 3, the method for removing invalid visitor trajectory fake big data of the present invention includes an actual trajectory data acquisition step (S100), a fake data generation step (S200), and a filtering step (S300).

More specifically, in the actual trajectory data acquisition step (S100), real trajectory data including content whose ID, time, and physical location in an arbitrary space are determined is acquired by the real trajectory data acquisition unit 100.

The ID referred to in the present invention may be at least one of unique numbers and letters provided to recognize individual visitors. The time referred to in the present invention is the time when a visitor arrives at a physical location in a certain space. Depending on one embodiment, time may include units of years, months, days, hours, minutes, and seconds. The arbitrary space may be implemented as two-dimensional coordinates, and thus the physical location of content in the arbitrary space may be specified. In other words, the order of content can be confirmed according to a certain direction. And the real trajectory data (real data) is data actually obtained from a large number of visitors in any space where exhibitions that visitors can view, such as exhibitions or science museums, are held. An example of the real trajectory data is shown in [Table 1].

In addition, the actual trajectory data acquisition step (S100) is characterized in that the contents of the actual trajectory data are arranged in time order by ID. This is because, when separately acquired data is listed in the order in which it was acquired, it is difficult to check the order and number of time-dependent content by ID, and it takes a considerable amount of time to check this, which reduces data processing speed and efficiency. It is to improve.

Next, in the fake data generation step (S200), the fake data generator 200 uses a deep learning model to generate fake data having the same columns as the actual trajectory data.

At this time, the fake data generation step (S200) is characterized in that the deep learning model is a CTGAN (Conditional Tabular GAN) model. In general, the GAN model is a network that artificially generates data similar to actual trajectory data including discrete data by utilizing mutually adversarial discriminators and generators, and is a model that belongs to unsupervised learning in machine learning. In other words, it is a model that complements the fact that the existing GAN model cannot be applied to discrete data.

Next, in the filtering step (S300), the filtering unit 300 filters out fake data that cannot be physically matched based on the actual trajectory data among the fake data.

The actual trajectory data referred to in the present invention has a time (t) value according to content (C), which is a specific location in arbitrary space. The fake data generated through CTGAN, the deep learning model, from the fake data generation step (S200) may include invalid data that violates the characteristics of the trajectory data. If the fake time value corresponding to a specific position among the fake data properties is a value that cannot be generated when looking at the front-end correlation of the actual trajectory data, the filtering step (S300) determines that the fake data is invalid. It may be removed as it is judged to be fake data or fake data that cannot be physically matched.

Specifically, in the filtering step (S300), the time of any fake data is within the time range of any real trajectory data by the time filter module 310 in the filtering unit 300, and the content of any fake data is If it exists within the content range of any actual trajectory data, it is judged to be valid fake data. By the time standard judgment step (S310) and the content filter module 320 in the filtering unit 300, the random fake data If the content exists within the content of random real trajectory data and the time of the random fake data exists within the time range of random real trajectory data, it includes a content standard judgment step (S320) in which it is judged to be valid fake data. It is characterized by

Looking at the embodiment of FIG. 2, an arbitrary space may include all contents (C0-Cn) whose physical locations are determined. Additionally, in the actual trajectory data acquisition step (S100), actual trajectory data including a plurality of contents (C0-Cn) dependent on time (t0-tn) may be acquired targeting ID 1. That is, the actual trajectory data may include C0 content at time t0, C2 content at t1, C5 content at t2, Cn-1 content at time tn-1, and Cn content at tn. Additionally, in the fake data generation step (S200), fake data including a plurality of contents (Cy1-Cm) dependent on time (tx1-txm) may be generated targeting ID 1. Here, n and m are positive integers. That is, the fake data may include Cy1 content at time tx1 and Cy2 content at time tx2.

If the tx1 time is within the range from to to t1, Cy1 content can be C1 content, but cannot be one of C3 to Cn content. In other words, Cy1 content cannot be physically matched if it is one of C3 to Cn content. That is, according to one embodiment, the time reference determination step (S310) determines that the fake data is invalid if the Cy1 content at time tx1 is C5 and is fake data that cannot be physically matched. This allows the corresponding fake data to be removed. Therefore, in the time-based determination step (S310), whether or not the content matches can be determined based on time.

Also, if Cy2 content is within the range of C2 content to C5 content, tx2 time must exist within the t1 to t2 time range. In other words, if the tx2 time is outside the t1 to t2 range, it cannot be physically matched. That is, the content standard determination step (S320) is invalid if the tx2 time of the Cy2 content does not exist in t1 to t2 and is t5 according to one embodiment, and is fake data that cannot be physically matched. It is determined that the corresponding fake data may be removed. Therefore, in the content standard determination step (S320), whether or not time matches can be determined based on the content.

Therefore, according to the present invention, by generating fake data having the same column as the actual trajectory data and filtering the fake data containing at least one of time and content that cannot be physically matched to the fake data, the necessary information for exhibition operation is provided. It has a remarkable effect in processing and correcting big data that simulates the trajectories of numerous visitors.

That is, conventionally, real trajectory data (Real data) is separated into real training data (Real training data) and real test data (Real test data) and at least one of the Neural Network, Decision Tree, and SVM models is used. An artificial intelligence model (DNN 1) including one was learned.

The present invention uses real test data among valid synthetic data and actual trajectory data filtered from the filtering unit 300 to create a neural network, decision tree, and You can learn an artificial intelligence model (DNN 2) that includes at least one of the SVM models. Accordingly, DNN 2 learned according to an embodiment of the present invention has the remarkable effect of deriving more accurate output values than DNN 1 learned using a conventional learning method.

Embodiments may be implemented by hardware, software, firmware, middleware, microcode, hardware description language, or any combination thereof. When implemented as software, firmware, middleware, or microcode, program code or code segments that perform necessary tasks may be stored in a computer-readable storage medium and executed by one or more processors.

And aspects of the subject matter described herein may be described in the general context of computer-executable instructions, such as program modules or components that are executed by a computer. Typically, program modules or components include routines, programs, objects, and data structures that perform specific tasks or implement specific data types. Aspects of the subject matter described herein may be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media, including memory storage devices.

As described above, although the embodiments have been described with limited examples and drawings, various modifications and variations can be made by those skilled in the art from the above description. For example, the described techniques are performed in an order different from the described method, and/or the components of the described system, structure, device, circuit, etc. are combined or combined in a different form than the described method, or in a different configuration. Appropriate results may be achieved through substitution or substitution by elements or equivalents.

Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims described below.

100.. Actual trajectory data acquisition unit
200.. Fake data generation unit
300.. Filtering unit
310.. Time filter module
320.. Content filter module

Claims (5)

a real trajectory data acquisition unit that acquires real trajectory data including content whose ID, time, and physical location in an arbitrary space are determined;
a fake data generator that generates fake data having the same columns as the actual trajectory data using a deep learning model; and
A filtering unit for filtering fake data that cannot be physically matched with the actual trajectory data among the fake data. An apparatus for removing invalid visitor trajectory fake big data including a. According to clause 1,
The actual trajectory data acquisition unit,
A device for removing invalid visitor trajectory imitation big data, characterized in that the actual trajectory data is arranged in chronological order by ID. According to clause 1,
The fake data generator,
A device for removing invalid visitor trajectory imitation big data, characterized in that the deep learning model is a CTGAN (Conditional Tabular GAN) model. According to clause 1,
The filtering unit,
a time filter module that determines the fake data to be valid if the time of the random fake data is within the time range of the random real trajectory data and the content of the random fake data is within the content range of the random real trajectory data; and
Includes a content filter module that determines that the content of any fake data is valid if the content of the fake data exists within the content of the random real trajectory data and the time of the random fake data exists within the time range of the random real trajectory data. A device for removing invalid visitor trajectory imitation big data. A real trajectory data acquisition step in which real trajectory data including content whose ID, time, and physical location in an arbitrary space are determined is acquired by a real trajectory data acquisition unit;
A fake data generation step in which fake data having the same column as the actual trajectory data is generated by a fake data generator using a deep learning model; and
A filtering step in which fake data that cannot be physically matched is filtered using the actual trajectory data as a standard among the fake data by the filtering unit; for removing invalid visitor trajectory fake big data including a method.